Running Back Projection Methodology: Workload and Efficiency Models
Running back projections sit at the intersection of two distinct measurement challenges: how often a player gets the ball, and what he does with it. The methodology covered here explains how workload metrics and efficiency models combine to produce per-game fantasy point estimates for NFL running backs — and where each model type breaks down under real-world conditions.
Definition and scope
A running back projection is a statistical forecast of expected fantasy scoring output, typically expressed as points-per-game under a specific scoring format. The projection integrates carry volume, target share, snap percentage, expected yards per attempt, and touchdown probability into a single estimate.
The workload side of the model covers opportunity — how many touches a back receives in a given game or week. The efficiency side covers production per touch — yards after contact, yards per route run, broken tackle rate, and similar per-touch metrics tracked by NFL data providers such as Pro Football Focus and Next Gen Stats (NFL, official tracking data).
Neither half of the model works well in isolation. A high-volume back on a bad offense can generate 20+ carries and still underperform projections if his efficiency metrics are bottom-quartile. A highly efficient back with a 35% snap share in a committee backfield faces a hard ceiling on total output regardless of how well he runs.
The scope of this methodology applies primarily to redraft and season-long fantasy formats. Dynasty-versus-redraft considerations shift some inputs — age curves and contract situations become relevant — but the core workload/efficiency framework transfers across contexts.
How it works
Running back projections are built in layers, typically in this sequence:
- Snap share baseline — The projected percentage of offensive snaps a back is expected to play, drawn from recent game logs and depth chart analysis. A back averaging 58% snap share over six games with no injury to the starter ahead of him carries a different baseline than a new starter promoted two weeks ago.
- Carry and target allocation — From the snap share baseline, models distribute expected carries and targets using team-level rushing and receiving volume projections. A team projecting 28 carries per game divided across a backfield splits differently than a single-back team.
- Yards-per-carry (YPC) and yards-per-reception (YPR) estimates — Efficiency inputs. These are regressed toward position-average benchmarks over time using the regression-to-mean framework to prevent early-season outliers from overweighting projections.
- Touchdown probability — A back's red zone carry share, combined with team-level touchdown rate, produces an expected touchdown contribution. This is frequently the highest-variance component in any single-game projection.
- Scoring format adjustment — PPR formats weight target volume more heavily; standard formats weight carry volume. Scoring format impact on projections details how the weighting shifts across league types.
- Contextual overlays — Vegas game totals, implied team scoring, weather conditions, and opponent run defense ranking all modify the base projection. See Vegas lines and fantasy projections and matchup-based projection adjustments for how these inputs are applied.
Common scenarios
The workhorse back — High snap share (65%+), high carry volume, limited receiving role. Projections here are relatively stable because workload variance is low. The risk is efficiency volatility: a back averaging 3.8 YPC over a four-game stretch will see projections adjust downward even if carries remain constant.
The committee back — Two or three backs splitting snaps roughly evenly. Projections for each back carry wider uncertainty bands because a single in-game injury or a halftime adjustment can redistribute 10+ carries. Projection confidence intervals explains how this uncertainty is quantified.
The receiving back — Lower carry volume but a 6-8 target-per-game profile. In PPR formats, this back's projection hinges almost entirely on target share, which is tracked through snap count and target share data. A 22% target share on a pass-heavy offense can produce RB2 projections despite 8 carries.
The handcuff — Projects as low-end starter or flex only if the lead back is active; projects as a high-upside starter if the lead back is injured. Models maintain two projection tracks for handcuffs, switching between them based on injury report status.
Decision boundaries
The workload/efficiency split creates specific thresholds where model behavior changes.
A snap share below 40% typically signals that a back lacks the volume floor to project as a reliable fantasy starter regardless of efficiency metrics. Above 60%, workload alone carries most of the projection weight, and efficiency inputs serve as refinement rather than primary drivers.
Efficiency metrics require a minimum of 40 carries before they stabilize enough to weight heavily — a principle consistent with sample size and projection reliability standards applied across position groups. Below that threshold, models lean more heavily on historical positional averages and pre-snap alignment data.
Touchdown regression is the most common decision boundary failure. A back who scored touchdowns on 8% of his red zone carries over three games will see that rate pulled toward the position average of roughly 3-4% (based on historical NFL red zone conversion data tracked by Pro Football Reference). Failing to apply that regression produces projections that are systematically overoptimistic.
The Fantasy Projection Lab home resource applies these layered inputs across the full NFL running back position, updating models as weekly usage data becomes available. For a full account of how projection outputs should be read and applied to lineup decisions, reading and interpreting projection outputs covers the translation from raw model numbers to actionable rankings.